An Improved Forecast Contest for 2010 -- your thoughts

M.T. Cranium

Ideas for the 2010 Forecast contest
____________________________________

In 2009, we have had a regular dozen entrants who seem to think the format being used is okay, but I believe we could have an even better contest and perhaps attract a larger field for 2010.

Here are some ideas I wanted to offer for discussion, as well as hearing any other ideas well in advance of the Christmas-New Years week when I will be away much of the time and needing to place the first month rules on the board.

First of all, let's review how the scoring is done. I give 30% of the points for a monthly forecast of temperature, taking a point off for every 0.1 error, and using the rather crude but somewhat effective average of highest and lowest monthly means reported in the monthly summary. This almost always turns out to be somewhere and Knock. Deep Easterly has been showing me that a more sophisticated system of the average of five selected stations gives a fairly similar value, but it has been somewhat lower for November. I am thinking of switching to this system for 2010, and entrants would be wise to think about the stations being used and how that might affect their guesses.

Secondly (and thirdly) I take the extremes for the month as reported in the summary, highest and lowest, and give each of them 25 points, losing one point per 0.1 error. This is easy and I probably won't switch this concept (but perhaps the scoring).

Then there are usually two bonus questions that I base on things that I figure might be interesting given the time of year or the set-up most are expecting. The knock against some of these has been too random, so in other words perhaps we should drop these bonus points in favour of more questions testing skill, such as monthly precip as percentage of normal, monthly sunshine, etc.

What I thought I might propose then, is this format which keeps one bonus question just for the fun of it.

(1) Predict the monthly temperature from DE's formula, which is announced in the contest notice (e.g., Feb 4.7 C is the long-term average) to give you some idea what to shoot for, then we have the advantage of getting regular updates of this value as the month proceeds. I could even ask for an anomaly given that there is a 30-year mean, but it makes no difference if we know what that 30-year mean value is. And I would reduce the points for this to 25, still using 0.1 as the error determinant.

(2) Monthly max, same as always, but points reduced to 20.

(3) Monthly min, same as always, but points reducted to 20.

(4) Monthly rainfall, percentage of normal, which I would read off the monthly summary from their easy-to-interpret graphs (thus using all reporting stations). This to be worth 15, and errors taken at 3% intervals for points.

(5) Monthly sunshine, percentage of normal, same as above, but just 10 points and using a 5% error interval.

(6) This leaves 10 points for a bonus question. I will continue with the usual assortment of mid-month basically guessing games but who knows there could be some skill involved in them.

Now, this leaves the question of deadlines for entry. I am more lenient with this boards contest than other weather forums which either lock the door at 2359h day before the month begins, or assign fairly harsh penalties on the first day. I don't really see the point of being too harsh since not much happens on the first day of the month that would "give away" the events of the rest of the month. But perhaps I should go to a system where any entry on the second loses ten points regardless (I have been saying 10% of what exceeds 60, and finding that I rarely have to apply that).

Bearing in mind that I want to keep this fun and easy to score, I look forward to your comments. By "easy to score" I need to have questions that I can verify from the monthly summary which is often available as early as the morning of the first of the month (very good work there by the way). Things that don't appear until the Bulletin, like snow lying days, etc, make for time-consuming enquiries to the met.ie staff, and I have probably used up my free request cards with them already (but must say they were very helpful when I had to get this info to score February 09).

I also need to have questions that suffer no ambiguity, such as "will it snow" questions, although I've given myself a few outs on that one for December.

This is why I like any kind of question that I can score from "yesterday's weather" in the met.ie website, such as strongest wind gust or daily max at a given site (so long as they are actually there that day).

Over to you then ....

(and yes, we would welcome new entrants for 2010, the idea is to score by monthly average points so that anyone can miss a month or two, say in the summer when you might be on holiday).

Mothman

M.T. Cranium wrote: »

Ideas for the 2010 Forecast contest
____________________________________


(the idea is to score by monthly average points so that anyone can miss a month or two, say in the summer when you might be on holiday).

I suggest taking 10 best months so that someone who partakes in all months can drop their 2 worst scores.

Deep Easterly

Just to expand on the method I put forward to M.T, it is pretty similar to the one I used a couple of year's ago when the competition started up, only this time using the combined means of 5 Met Eireann stations as oppossed to the 10 that were used originally.

Out of personal interest, I like to keep a record of daily temperature and rainfall figures from around Ireland simply because I have an unhealthy obsession with all things weather and how this weather effects our wee North Atlantic rock.

Because of time constraints and so forth, I narrowed down my daily record keeping to just 5 stations which are reasonably well postioned around Ireland. All 5 are placed inland and as M.T rightly pointed out, the combined means of those stations equated to a slightly lower value than what was shown by his own method. This is simply just to do with the time of year, because in summer period, the method I use would probably show a slightly higher value than might be shown in M.T's version, so overall, there would be little, if any, difference on a yearly basis between between the two.

The method I put forward would use the daily means of 5 met eireann stations as I have already mentioned, to add up to a monthly mean which would be compared to the 61-90 average mean of them and give us an idea of what to expect on a montly basis with regards both temperature and rainfall.

The 5 Met Eireann stations I propose to use are as folllows:

Casement: being in the east, it would represent the eastern region nicely, as it would be the driest station of the 5.

Claremorris: this station would give a good idea of rainfall and temperature over the western region.

Mullingar: would represent the north midland and northeastern region and is one of the coolest stations of the 5 on a yearly basis.

Oak Park: would give an good idea of what is happening in the south midland and southeastern region on a monthly basis.

Shannon Airport: The warmest of the 5 on a year in, year out basis, and would be fairly well representative of what is happening in the mid/southwest with regards temperature and rainfall.

Of course just using the combined means of both temp and rainfall of these 5 met eireann stations alone would not represent the profile of monthly means over Ireland as a whole, but it might be interesting to monitor all the same.

The above is just my suggestion for the competiton. If others have different views on what might be good for the competition then please let M.T. know before the end of the month. Whichever proposal is put forward, it should make for another interesting and exciting year for the comp!!:)

pauldry

Hello Guys,

Im fairly new to the whole weather board ( as you can see by the number of posts beside my name). When is the closing date for these predicting competitions and is there any prize or is it just a bit of craic? I just missed out on the December one. Also when are the results printed i.e. what is the name of the thread they are printed in?

M.T. Cranium

Just for fun, closing date has been generally the first of each month, and each month has its own thread, while scoring is shown in an "annual forecast contest scoring" thread. These should be down the first page somewhere at this point, or possibly top of page two of forum topics.

Deep Easterly

Have been experimenting today with historic daily means for December for Claremorris. Taking daily data for this month over a 35 year period from this site:

http://www.tutiempo.net/en/Climate/CLAREMORRIS/39700.htm

it was possible to come up with this 35 year daily mean chart:



which discludes data between 1996 and 1999 as none was available for the station.

This chart would give an idea only, of average temperature trends thoughout each month. Naturally, I would include data from the other stations that may be used in next year's comp, although for Oak Park I would have to use data from the now defunct Kilkenny.

Just an idea at the moment.

M.T. Cranium

I haven't heard much, which I assume means that people are okay with the suggested format above. But I'll wait until the weekend of the 18th-19th before closing discussion, so if anyone wants to add their thoughts, try to get them in here by then.

DE, if you are amenable, I would go with your five-station average as our new temperature standard, as long as you can provide a target figure around the 27th of each month when I open the new monthly competition. When you give this value (the long-term average at those five stations) then it gives our contestants a point of reference. That Claremorris data would be interesting as an aside to the contest and for updates during the month.

Otherwise, I think I'll go with rainfall and sunshine averages that I can calculate from the met.ie website, and so the only work you would need to do would be to supply us with the target figure for the coming month, and whatever updates you wanted to post in the thread as the month goes by.

Thanks a lot for your offer to improve our contest, I hope indeed it will do that, along with the other proposed changes.

octo

M.T. Cranium wrote: »

(4) Monthly rainfall, percentage of normal, which I would read off the monthly summary from their easy-to-interpret graphs (thus using all reporting stations). This to be worth 15, and errors taken at 3% intervals for points.

(5) Monthly sunshine, percentage of normal, same as above, but just 10 points and using a 5% error interval.

Good Work Lads - Just a few remarks

Just to be clear, the average rainfall and s/s are averages of the data posted in the monthly weather summaries, and will be one figure for the month?

Btw, my understanding is that Malin Head and Belmullet are soon to be replaced with automatic weather stations, and there won't be any sunshine values as there won't be anyone to change the card on the sunshine recorder. (Correct me someone if I'm wrong)

It seems a pity not to include Valentia - its the oldest and most climatalogically significant met station in the country.

M.T. Cranium

Yes, the contest will ask for one average value for each element, as well as the two temperature extremes. If the summary changes the menu of stations, we'll have to go along with what they publish. DE has five stations selected for the temperature, if he leaves out Valentia and Knock that will probably cancel out, from what I've seen these are usually the two extremes, and the system we were using in 2009 was the average of the two highest and lowest, which usually meant those two. We've compared notes and the adjustment will not make much difference to that temperature value.

Anyway, hope you'll all join in the fun (stressing fun) -- we've had a good close competition in 2009 that is still anyone's to win (well almost anyone) and Danno holding a slight lead.

I will post the January competition fairly early compared to most months, because of the holiday period, so DE if you could let me know what to say about January temperature, I will start thinking about the bonus question.